Qwen90M step 770000; complete resumable checkpoint

complete.json

@@ -1,18 +1,18 @@
 {
-  "step": 765000,
+  "step": 770000,
   "files": {
     "config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7",
     "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4",
-    "model.safetensors": "6b4ad24005ad10d830f5b4dd0b826962f001a14e87dcbc136addc786b07770c8",
+    "model.safetensors": "e0a55fb5bc214d02022f2eff8a8e4d4a8989e853edc73e1a2ebb5d0b7571c743",
     "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03",
     "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300",
     "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263",
     "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6",
-    "trainer.pt": "07dc267a3de4e00552f0a06fb17ec8e0dcd6e1e709ac7d235e0e71960df8e5ea",
+    "trainer.pt": "a498688babdf2cae771028b6c89d36575e2ed244953a97468586e9e5595fe80e",
     "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd",
     "performance.json": "d91455eeac02a2960a60e3a2edc7d99796576b62edb422642a8beb3b824ff02d",
-    "shard_scheduler.json": "4c919b423fdadb08dc04e9ee2624261806d3f9f3f73d63e1a0959bb729e52ff3",
-    "coverage.json": "b421dc96b600cc122bd4c40c0c3b6e29b4666eec433d9dfb0299a93825a0d58c",
+    "shard_scheduler.json": "d4f4b8859106eed29ba6de07b880ae2a733ce84e3ecc5289a12c47202ed328b5",
+    "coverage.json": "0b5c9661fa1c47243f8665233277a741ea19c000bbb01a96a0cb91fec8b18598",
     "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"
   },
   "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e"

coverage.json

@@ -1,107 +1,107 @@
 {
-  "step": 765000,
+  "step": 770000,
   "stats": {
-    "climbmix/blocks": 27540000,
-    "climbmix/targets": 46037548584,
-    "climbmix/input_tokens": 46040465939,
-    "ultra_qa/blocks": 10327500,
-    "ultra_qa/targets": 16848102853,
-    "ultra_qa/input_tokens": 16848135318,
-    "cortex/blocks": 10327500,
-    "cortex/targets": 392690448,
-    "cortex/input_tokens": 16416010630,
-    "ultra_style/blocks": 10327500,
-    "ultra_style/targets": 16334987427,
-    "ultra_style/input_tokens": 16335844116,
-    "openmath/blocks": 6885000,
-    "openmath/targets": 10638497464,
-    "openmath/input_tokens": 10642041928,
-    "rewrite6/blocks": 3442500,
-    "rewrite6/targets": 5643572276,
-    "rewrite6/input_tokens": 5649759720
+    "climbmix/blocks": 27720000,
+    "climbmix/targets": 46338195596,
+    "climbmix/input_tokens": 46341132368,
+    "ultra_qa/blocks": 10395000,
+    "ultra_qa/targets": 16958238643,
+    "ultra_qa/input_tokens": 16958271297,
+    "cortex/blocks": 10395000,
+    "cortex/targets": 395254793,
+    "cortex/input_tokens": 16523352599,
+    "ultra_style/blocks": 10395000,
+    "ultra_style/targets": 16442103528,
+    "ultra_style/input_tokens": 16442965847,
+    "openmath/blocks": 6930000,
+    "openmath/targets": 10708073683,
+    "openmath/input_tokens": 10711641641,
+    "rewrite6/blocks": 3465000,
+    "rewrite6/targets": 5680408647,
+    "rewrite6/input_tokens": 5686636248
   },
   "sources": {
     "climbmix": {
-      "page": 3771,
-      "row": 3176,
+      "page": 3796,
+      "row": 2700,
       "start": {
-        "file": 182,
+        "file": 183,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 44,
+          "row_group": 61,
           "row": 1024
         }
       },
       "epoch": 0,
-      "offset": 14336,
-      "documents": 15449192
+      "offset": 2048,
+      "documents": 15551116
     },
     "rewrite6": {
-      "page": 45,
-      "row": 3598,
+      "page": 9,
+      "row": 2929,
       "start": {
         "file": 0,
         "reader": {
           "format": "jsonl-row-v1",
-          "row": 184320
+          "row": 36864
         }
       },
-      "epoch": 42,
+      "epoch": 43,
       "offset": 0,
-      "documents": 8814886
+      "documents": 8872165
     },
     "openmath": {
-      "page": 2357,
-      "row": 768,
+      "page": 2372,
+      "row": 2873,
       "start": {
         "file": 22,
         "reader": {
           "format": "parquet-row-v1",
-          "row_group": 47,
-          "row": 972
+          "row_group": 109,
+          "row": 412
         }
       },
       "epoch": 0,
       "offset": 0,
-      "documents": 9655040
+      "documents": 9718585
     },
     "ultra_qa": {
-      "page": 1019,
-      "row": 952,
+      "page": 1025,
+      "row": 3695,
       "start": {
         "file": 8,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 0,
-          "row": 16813
+          "row": 41389
         }
       },
       "epoch": 0,
-      "offset": 2048,
-      "documents": 4174776
+      "offset": 0,
+      "documents": 4202095
     },
     "ultra_style": {
-      "page": 1555,
-      "row": 2608,
+      "page": 1565,
+      "row": 3159,
       "start": {
         "file": 9,
         "reader": {
           "format": "parquet-row-v1",
           "row_group": 2,
-          "row": 34550
+          "row": 75510
         }
       },
       "epoch": 0,
       "offset": 2048,
-      "documents": 6371888
+      "documents": 6413399
     },
     "cortex": {
-      "page": 200288,
-      "row": 88,
+      "page": 201598,
+      "row": 95,
       "start": null,
       "epoch": 0,
       "offset": 0,
-      "documents": 25636952
+      "documents": 25804639
     }
   }
 }

latest.json

@@ -1 +1 @@
-{"step": 765000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "6b4ad24005ad10d830f5b4dd0b826962f001a14e87dcbc136addc786b07770c8", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "07dc267a3de4e00552f0a06fb17ec8e0dcd6e1e709ac7d235e0e71960df8e5ea", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "d91455eeac02a2960a60e3a2edc7d99796576b62edb422642a8beb3b824ff02d", "shard_scheduler.json": "4c919b423fdadb08dc04e9ee2624261806d3f9f3f73d63e1a0959bb729e52ff3", "coverage.json": "b421dc96b600cc122bd4c40c0c3b6e29b4666eec433d9dfb0299a93825a0d58c", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000765000"}
+{"step": 770000, "files": {"config.json": "cbc72137da3dae65085a2a8c39ad7634d16bf92d785ca00d283ee82fe32b37e7", "generation_config.json": "54ed0d19f6226888c91157964ede8a892bd5ae175bfad85336152b3d9b6cd7c4", "model.safetensors": "e0a55fb5bc214d02022f2eff8a8e4d4a8989e853edc73e1a2ebb5d0b7571c743", "tokenization_byte.py": "279b9990f77b1eaedc7c1184fe13d60273a713ead856200f1fb9ae323c7b1e03", "tokenizer_config.json": "0ecbd4fd39e35125800d806af981ba0cacd8a7ba5f02c9e729c4ecc54af6d300", "special_tokens_map.json": "20d4b94d0e657c3f44d9123522c106e7315b6c1a1b6d259478f17bd2918b1263", "byte_vocab.json": "57a1cd2e70dd5b7cc6fd2eaa12ebfcd173c1620fa5d45c2c6282511c29b6a5f6", "trainer.pt": "a498688babdf2cae771028b6c89d36575e2ed244953a97468586e9e5595fe80e", "training_config.json": "4e8f8b2fc7193aabdbca12082e7c7cc3a0ed080bd5ff16e060cc02e10ffc53dd", "performance.json": "d91455eeac02a2960a60e3a2edc7d99796576b62edb422642a8beb3b824ff02d", "shard_scheduler.json": "d4f4b8859106eed29ba6de07b880ae2a733ce84e3ecc5289a12c47202ed328b5", "coverage.json": "0b5c9661fa1c47243f8665233277a741ea19c000bbb01a96a0cb91fec8b18598", "README.md": "9ce0f9efacbc46177a7e818b6aed274d84e7145ff944b5c7d6ecd5d33b9d4419"}, "signature": "052aedaef69b550457da29474e17df28fa635c99594a7b9c0881f3ac3827000e", "folder": "step-000000770000"}

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:6b4ad24005ad10d830f5b4dd0b826962f001a14e87dcbc136addc786b07770c8
+oid sha256:e0a55fb5bc214d02022f2eff8a8e4d4a8989e853edc73e1a2ebb5d0b7571c743
 size 360622744

shard_scheduler.json

@@ -22,8 +22,8 @@
       "tokens_per_second": 598315.063954628,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 77160.6514700967,
-      "required_pages_per_second_upper_estimate": 0.0053381529060792365
+      "required_token_positions_per_second": 76400.702468659,
+      "required_pages_per_second_upper_estimate": 0.005285577870835155
     },
     {
       "source": "rewrite6",
@@ -36,8 +36,8 @@
       "tokens_per_second": 8533.957213384612,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 9645.081433762087,
-      "required_pages_per_second_upper_estimate": 0.004038194184112503
+      "required_token_positions_per_second": 9550.087808582375,
+      "required_pages_per_second_upper_estimate": 0.003998422336942226
     },
     {
       "source": "openmath",
@@ -50,8 +50,8 @@
       "tokens_per_second": 380144.35761751566,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 19290.162867524174,
-      "required_pages_per_second_upper_estimate": 0.004017632209500813
+      "required_token_positions_per_second": 19100.17561716475,
+      "required_pages_per_second_upper_estimate": 0.003978062875551665
     },
     {
       "source": "ultra_qa",
@@ -64,8 +64,8 @@
       "tokens_per_second": 43742.16556775435,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 28935.244301286257,
-      "required_pages_per_second_upper_estimate": 0.0017346237143022584
+      "required_token_positions_per_second": 28650.263425747125,
+      "required_pages_per_second_upper_estimate": 0.0017175395459542889
     },
     {
       "source": "ultra_style",
@@ -78,8 +78,8 @@
       "tokens_per_second": 22018.86276523721,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 1,
-      "required_token_positions_per_second": 28935.244301286257,
-      "required_pages_per_second_upper_estimate": 0.0022495879717359174
+      "required_token_positions_per_second": 28650.263425747125,
+      "required_pages_per_second_upper_estimate": 0.002227431962161148
     },
     {
       "source": "cortex",
@@ -92,8 +92,8 @@
       "tokens_per_second": 910315.184299418,
       "note": "Bounded fresh sample; full-page latency is an extrapolation.",
       "estimated_pages_for_buffer": 23,
-      "required_token_positions_per_second": 28935.244301286257,
-      "required_pages_per_second_upper_estimate": 0.37115024565213706
+      "required_token_positions_per_second": 28650.263425747125,
+      "required_pages_per_second_upper_estimate": 0.3674948169693453
     }
   ],
   "gpu_measurement": "Selected forward/backward estimate; replaced by live successful update timings.",
@@ -158,24 +158,24 @@
     "prefetch_batches": 63,
     "prefetch_capacity_limit": 512,
     "batch_memory_estimate_bytes": 720750,
-    "producer_batch_seconds": 0.22211623922648727,
-    "producer_p95_seconds": 0.16718530399975862,
-    "gpu_update_seconds": 0.9555129278369688,
-    "producer_batches_per_second": 4.5021471797040515,
-    "gpu_batches_per_second": 1.0465583152953652,
-    "producer_headroom": 4.30185983323197,
+    "producer_batch_seconds": 0.0745995468515872,
+    "producer_p95_seconds": 0.19039384700045048,
+    "gpu_update_seconds": 0.9650173050469608,
+    "producer_batches_per_second": 13.40490716370516,
+    "gpu_batches_per_second": 1.036250847285414,
+    "producer_headroom": 12.935967385523451,
     "sustainable": true,
-    "buffer_seconds": 60.197314453729035,
+    "buffer_seconds": 60.79609021795853,
     "buffer_drain_seconds": null,
     "shard_target_mib": 16,
-    "upload_pages_per_commit": 79,
+    "upload_pages_per_commit": 84,
     "max_pending_pages": 4096,
-    "generation_bytes_per_second": 471718.0651086896,
-    "upload_bytes_per_second": 5138876.858680128,
+    "generation_bytes_per_second": 277550.1790140697,
+    "upload_bytes_per_second": 5458592.488142992,
     "upload_sustainable": true,
     "logical_page_rows": 4096,
     "logical_cortex_page_rows": 128,
-    "estimated_full_shards_per_hour": 101.21971573777691,
+    "estimated_full_shards_per_hour": 59.555807378926936,
     "selection_policy": "Smallest shard within 90% of measured local peak, enlarged to amortize 30 seconds of measured page production; bounded to 256 MiB."
   },
   "upload_measurements": [
@@ -244,6 +244,252 @@
       "bytes": 12097526,
       "pages": 37,
       "shards": 3
+    },
+    {
+      "seconds": 0.8632086769994203,
+      "bytes": 2602865,
+      "pages": 34,
+      "shards": 2
+    },
+    {
+      "seconds": 1.1280111480000414,
+      "bytes": 298657,
+      "pages": 9,
+      "shards": 1
+    },
+    {
+      "seconds": 1.3505803890002426,
+      "bytes": 6226670,
+      "pages": 36,
+      "shards": 2
+    },
+    {
+      "seconds": 1.6309672779998436,
+      "bytes": 11574403,
+      "pages": 34,
+      "shards": 4
+    },
+    {
+      "seconds": 1.1955943939992721,
+      "bytes": 921426,
+      "pages": 30,
+      "shards": 1
+    },
+    {
+      "seconds": 1.0380561770007262,
+      "bytes": 6122044,
+      "pages": 36,
+      "shards": 2
+    },
+    {
+      "seconds": 1.4052477490004094,
+      "bytes": 13455700,
+      "pages": 38,
+      "shards": 4
+    },
+    {
+      "seconds": 0.9797629530003178,
+      "bytes": 4961107,
+      "pages": 36,
+      "shards": 2
+    },
+    {
+      "seconds": 1.0490742649999447,
+      "bytes": 7239835,
+      "pages": 30,
+      "shards": 3
+    },
+    {
+      "seconds": 1.0533643100006884,
+      "bytes": 1118487,
+      "pages": 34,
+      "shards": 1
+    },
+    {
+      "seconds": 1.5156246349997673,
+      "bytes": 10630930,
+      "pages": 34,
+      "shards": 3
+    },
+    {
+      "seconds": 1.1701987970000118,
+      "bytes": 7151255,
+      "pages": 34,
+      "shards": 3
+    },
+    {
+      "seconds": 1.1764370720002262,
+      "bytes": 6881233,
+      "pages": 35,
+      "shards": 2
+    },
+    {
+      "seconds": 1.185791879000135,
+      "bytes": 7588973,
+      "pages": 38,
+      "shards": 3
+    },
+    {
+      "seconds": 0.8400346700000227,
+      "bytes": 806070,
+      "pages": 24,
+      "shards": 1
+    },
+    {
+      "seconds": 1.260675972000172,
+      "bytes": 10735591,
+      "pages": 41,
+      "shards": 3
+    },
+    {
+      "seconds": 1.1190538959999685,
+      "bytes": 7646532,
+      "pages": 38,
+      "shards": 3
+    },
+    {
+      "seconds": 1.0387209649998113,
+      "bytes": 1064900,
+      "pages": 32,
+      "shards": 1
+    },
+    {
+      "seconds": 1.1239516590003404,
+      "bytes": 11388095,
+      "pages": 24,
+      "shards": 3
+    },
+    {
+      "seconds": 1.270762184999512,
+      "bytes": 12213784,
+      "pages": 48,
+      "shards": 4
+    },
+    {
+      "seconds": 1.305295600000136,
+      "bytes": 1053111,
+      "pages": 30,
+      "shards": 1
+    },
+    {
+      "seconds": 1.13043969499995,
+      "bytes": 7715241,
+      "pages": 34,
+      "shards": 3
+    },
+    {
+      "seconds": 1.1387970109999515,
+      "bytes": 10453428,
+      "pages": 40,
+      "shards": 3
+    },
+    {
+      "seconds": 1.3777619700003925,
+      "bytes": 819264,
+      "pages": 26,
+      "shards": 2
+    },
+    {
+      "seconds": 1.2043475289992784,
+      "bytes": 13122126,
+      "pages": 40,
+      "shards": 4
+    },
+    {
+      "seconds": 1.1971378379994349,
+      "bytes": 5971358,
+      "pages": 36,
+      "shards": 2
+    },
+    {
+      "seconds": 0.9979249650004931,
+      "bytes": 6639528,
+      "pages": 32,
+      "shards": 3
+    },
+    {
+      "seconds": 1.0738694510000641,
+      "bytes": 6116466,
+      "pages": 37,
+      "shards": 2
+    },
+    {
+      "seconds": 0.993601779000528,
+      "bytes": 1093689,
+      "pages": 34,
+      "shards": 1
+    },
+    {
+      "seconds": 1.0732885580000584,
+      "bytes": 7199351,
+      "pages": 29,
+      "shards": 3
+    },
+    {
+      "seconds": 1.3326483809996716,
+      "bytes": 16444186,
+      "pages": 37,
+      "shards": 4
+    },
+    {
+      "seconds": 0.9647154859994771,
+      "bytes": 2573191,
+      "pages": 37,
+      "shards": 2
+    },
+    {
+      "seconds": 1.2671719289992325,
+      "bytes": 6070484,
+      "pages": 36,
+      "shards": 2
+    },
+    {
+      "seconds": 0.9431054610013234,
+      "bytes": 842654,
+      "pages": 26,
+      "shards": 1
+    },
+    {
+      "seconds": 1.3556453979999787,
+      "bytes": 11128099,
+      "pages": 32,
+      "shards": 4
+    },
+    {
+      "seconds": 1.1991763389996777,
+      "bytes": 5989028,
+      "pages": 39,
+      "shards": 2
+    },
+    {
+      "seconds": 1.483354228999815,
+      "bytes": 13516314,
+      "pages": 40,
+      "shards": 4
+    },
+    {
+      "seconds": 0.855230520999612,
+      "bytes": 838138,
+      "pages": 29,
+      "shards": 1
+    },
+    {
+      "seconds": 1.1489447340009065,
+      "bytes": 10403004,
+      "pages": 36,
+      "shards": 3
+    },
+    {
+      "seconds": 0.9511843500004034,
+      "bytes": 2374318,
+      "pages": 29,
+      "shards": 2
+    },
+    {
+      "seconds": 1.678345239999544,
+      "bytes": 6141170,
+      "pages": 36,
+      "shards": 2
     }
   ],
   "cache_key": {

trainer.pt

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:07dc267a3de4e00552f0a06fb17ec8e0dcd6e1e709ac7d235e0e71960df8e5ea
-size 721777663
+oid sha256:a498688babdf2cae771028b6c89d36575e2ed244953a97468586e9e5595fe80e
+size 721763327